This paper is focused on dynamic coverage control with a team of robots. In this framework, decentralized control algorithms have been investigated to deal with the efficient coordination of the resources. The main contribution is a novel global strategy based on a hierarchical grid decomposition of the domain. This decomposition allows an ordered coverage of the domain that combined with a gradient based control law of the local error, achieves a better performance than previous approaches of dynamic coverage. The total coverage of the domain is proven, and the good performance of the approach is supported with simulations.